[Question] How do you analyzed PPG-BP dataset with small duration signal?

[almirb]

Hi!
I was trying to get signal analysis based on records of PPG-BP dataset, but when I run s = PPG(signal), I always get the message AssertionError: Signal must be at least fifteen seconds .

Looking on results available for download in this page, I noticed some files like this one:

Here we can see manual annotations (I think) and pyPPG detected fiducials, on an 2.1 seconds file.

I also tried to disable the ppg_len checking (s = PPG(signal, check_ppg_len=False)), and the code works. But when running fiducials = fpex.get_fiducials(s) , I got the error: AttributeError: 'list' object has no attribute 'astype'.

How can I get those fiducial points working as you did on PPG-BP dataset?

P.S.: On this issue, the user tiled the signal as much was needed to make the analysis.... and I don't know if it is a good practice... (honestly, I think isn't, but I don't have a better idea).

Thanks!

[gpwhs]

I would also like some info on this, I seem to remember it working before for single waveform signals when using check_ppg_len=False, but now I'm trying to use the package again it appears there are quite a few changes/inconsistencies between the release version available via pip and the methods used in the example on github.

For example, the PPG constructor only accepts s as an argument now and check_ppg_len is nowhere to be found, and the preprocessing API has completely changed in the release version when compared to the source available here https://github.com/godamartonaron/GODA_pyPPG/blob/d826dc3c61950c99df6b86afe8a0027e3c559280/pyPPG/preproc.py

@godamartonaron @peterhcharlton any input here?

[godamartonaron]

Hi,
Apologies for the late response. Due to a fiducial point detection algorithm, a 15-second PPG signal is required. For the BBP-BP dataset analysis, we have developed a special validation module.
If you want to analyze a single pulse wave, please use this module:
GitHub - GODA_pyPPG Validation Module
I hope this helps!
Best regards,
Marton

[gpwhs]

@almirb If you're still looking for how short PW biomarker extraction works (e.g. 1 waveform), I have a fork where I implemented a (very!) hacky way of doing it. It's specific to my use case right now, but should be helpful. It's in /validation/pw_anal.py. I also sped up the extraction process for large numbers of signals as it was very slow for me once it got above ~1500 signals:
https://github.com/gpwhs/goda_pyppg/

Hope this is okay @godamartonaron :)

[godamartonaron]

Thank you, @gpwhs !